FORMAL SYSTEMS

EigenForm
Louis H. Kauffman
University of Illinois, Chicago, Illinois, USA
Abstract
Purpose Discusses the notion of eigenform as explicated by Heinz von Foerster wherein an object
is seen to be a token for those behaviors that lend the object its apparent stability in a changing world.
Design/methodology/approach Describes von Foersters model for eigenforms and recursions
and put this model in the context of mathematical recursions, fractals, set theory, logic, quantum
mechanics, the lambda calculus of Church and Curry, and the categorical framework of xed points of
Lawvere.
Findings Determines that iterating an object upon itself is seen to be a key to understanding the
nature of objects and the relationship of an observer and the apparent world of the observer.
Originality/value Contemplates the concept of recursion in the context of second-order
cybernetics.
Keywords Cybernetics, Philosophical concepts, Mathematical sciences
Paper type Conceptual paper
1. Introduction
This essay discusses the notion of eigenform as explicated by von Foerster, 1981a).
A strong source of Heinzs work on eigenforms and self-reference is his earlier
consideration of second-order cybernetics in the form of cybernetics of cybernetics.
We consider the possibility of a eld of intellectual endeavor that can investigate itself,
and the way in which such a turn can come about. Let us begin by listing some of these
turns inward:
.
Cybernetics of Cybernetics
.
Mathematics of Mathematics
.
Computation of Computation
.
Linguistics of Linguistics
.
Magic of Magic
.
Logic of Logic
.
Geometry of Geometry
.
Pattern of Pattern
.
Will of Will
The Emerald Research Register for this journal is available at The current issue and full text archive of this journal is available at
www.emeraldinsight.com/researchregister www.emeraldinsight.com/0368-492X.htm
The author takes great pleasure to thank Annetta Pedretti, Ranulph Glanville, Nathaniel
Hellerstein and James Flagg for their conversations genetic to this paper. This paper is an
extended version of an essay on the same topic in the journal Cybernetics and Human Knowing
(Kauffman, 2003). The author wishes to thank Soren Brier for many conversations that
sharpened the content of both papers, and the referees for this paper whose comments were quite
helpful in clarifying a number of points.
K 7137721/12/2004RAVICHANDRAN127841
EigenForm
129
Received July 2003
Accepted February 2003
Kybernetes
Vol. 34 No. 1/2, 2005
pp. 129-150
qEmerald Group Publishing Limited
0368-492X
DOI 10.1108/03684920510575780
.
Teaching of Teaching
.
Learning of Learning
.
Meaning of Meaning
Heinz performs the magic trick of convincing us that the familiar objects of our
existence can be seen to be nothing more than tokens for the behaviors of the organism
that apparently create stable forms. These stabilities persist, for that organism, as an
observing system. This is not to deny an underlying reality that is the source of
objects, but rather to emphasize the role of process, and the role of the organism in the
production of a living map, a map that is so sensitive that map and territory are
conjoined. Von Foersters (1981a-d) book and papers were instrumental in pioneering
the eld of second-order cybernetics.
The notion of an eigenform is inextricably linked with second-order cybernetics.
One starts on the road to such a concept as soon as one begins to consider a pattern of
patterns, the form of form or the cybernetics of cybernetics. Such concepts appear to
close around upon themselves, and at the same time they lead outward. They suggest
the possibility of transcending the boundaries of a system from a locus that might have
been within the system until the circular concept is called into being. But then the
boundaries have turned inside out, and the inside is the outside.
Forms are created from the concatenation of operations upon themselves and objects
are not objects at all, but rather indications of processes. Upon encountering an object as
such a form of creation, you are compelled to ask: How is that object created? How is it
designed? What do I do to produce it? What is the network of productions? Where is the
home of that object? In what context does it exist? How am I involved in its creation?
Taking Heinzs suggestion to heart, we nd that an object is a symbolic entity,
participating in a network of interactions, taking on its apparent solidity and stability
from these interactions. We ourselves are such objects, we as human beings are signs
for ourselves, a concept originally due to the American philosopher C. S. Peirce
(Kauffman, 2001b). In many ways, Heinzs eigenforms are mathematical companions to
Peirces work. We will not follow this comparison in the present essay, but the reader
familiar with Peirce is encouraged to do so.
The paper is organized as follows. Sections 2 and 3 are a discussion of the nature of
object as we tend to describe it in language, leading from the descriptions and
assumptions that we use in the everyday world to the somewhat different concepts
of object that occur in scientic work, particularly in the physics of the very small.
We indicate how the concept of object arising as a token for the interaction and mutual
production of the observer, and the observed intermediates among these points of view.
Section 4 gives a short exposition of the mathematical form of Heinzs model. Once this
mathematics of eigenforms is on the table, one can discuss how it is related to our use
and concept of the notion object.
Section 5 points out just how the von Foerster eigenform and the eigenvector
of quantum physics are related. We give some hints about using eigenforms to
understand quanta. The view of physics from the eigenforms is a reversal of
epistemology, a complete turning of the world upside down. Heinz has tricked us into
considering the world of our experience and nding that it is our world, generated by
our actions. The world has become apparently objective through the self-generated
stabilities of those very actions.
K 7137721/12/2004RAVICHANDRAN127841
K
34,1/2
130
Section 6 is a recounting of a conversation of the author and Ranulph Glanville in
which Ranulph asked Does every recursion have a xed point? In the language of the
present essay this is the question Does every process have an eigenform?.
The obvious answer is no, but the answer that comes from Heinzs model is yes! There
is always an ideal eigenform. The challenge is to integrate that form into the context of
ones living.
In Sections 7-9 we consider questions about eigenforms in relation to mathematical
objects such as numbers and linguistic objects such as names.
Section 10 describes how to construct eigenforms without an excursion to innity,
by the xed point method of the logicians Church and Curry. We point out that
the construction of eigenforms in the sense of Heinzs model can be done without the
idealized excursion to innity (used by Heinz) via a device invented by Church and
Curry in the 1930s that is commonly called the lambda calculus. This device rst
involves constructing an intermediate operator (that I like to call the gremlin) who
acts as a catalyst for a process. When gremlins meet they produce eigenforms!
Section 11 shows how differentiation in formal calculus is related to the inductive
construction of number, and to eigenforms for the derivative. We discuss numbers as
eigenforms, and here it will become immediately apparent that with number one as a
perfect example of how an eigenform is both an object and an operator, how it is both
specic and ideal. This harks back to our discussion of quantum mechanics and
eigenforms.
In Sections 11 and 12 we have created a model for a set theoretic world that contains
a parable of the mental and the physical. The purely mental world is the class of sets
generated from the empty set. The purely physical world is the class of sets generated
from the special objects a that are eigenforms for the operation of framing, so that
a = {a}: The interface of mental and physical occurs as these realms touch in the limit
of nested parentheses
W = {{{{{{{. . .}}}}}}}:
where W is an amphibian living in both worlds. W is the eigenform that crosses the
boundary from the mental to the physical.
In Section 14 we consider naming and self-reference and return to Heinzs denition
of I. We end Section 14 with a discussion of how self-reference occurs in language,
through an indicative shift that welds the name of a person to his/her physical presence
and shifts the indication of that name to a metaname. More could be said at this point,
as the indicative shift is a linguistic entry into the world of Godelian sentences and the
incompleteness of formal systems. We emphasize the natural occurrence of eigenforms
in the world of our linguistic experience and how this occurrence is intimately
connected to our structure as self-observing systems.
In Section 15 we show how eigenforms and imaginary Boolean values emanate
naturally from the category theorist William Lawveres amazing proof of Cantors
theorem on subsets of an innite set. The paper ends with a short epilogue in
Section 16.
2. Objects
What is an object? At rst glance, the question seems perfectly obvious. An object is a
thing, i.e. something that you can pick up and move and manipulate in
K 7137721/12/2004RAVICHANDRAN127841
EigenForm
131
three-dimensional space. An object is three-dimensional, palpable, like an apple or a
chair, or a pencil or a cup. An object is the simplest sort of entity that can be subjected
to reference. All language courses rst deal with simple objects like pens and tables.
La plume est sur la table.
An object is separate from me. It is out there. It is part of the reality separate from
me. Objects are composed of objects, their parts. My car is made of parts. The chair is a
buzzing whirl of molecules. Each molecule is a whirl of atoms and each atom a little
solar system of electrons, neutrons and protons. But wait! The nucleus of the atom is
composed of strange objects called quarks. No one can see them. They do not exist as
separate entities. The electrons in the atom are special objects that are not separate
from each other and from everything else. And yet when you observe the electrons,
they have denite locations.
The physicists world divides into quantum objects that are subject to the
constraints of the uncertainty principle, and classical objects that live in the dream of
objective existence, carrying all their properties with them.
A classical object has a location at a given time. You can tell where it is. You can tell
a story of where it has been. If the classical object breaks up into parts, you will be able
to keep track of all the parts. Yet, electrons and positrons can meet each other and
disappear into pure energy! Should we allow objects to disappear? What sort of an
object is the electromagnetic eld of radio and television signals that oods this room?
Is my thought to be thought of as an object? Can I objectify my thought by writing it
down on paper or in the computer? Am I myself an object? Is my body an object in the
three-dimensional space? Is the space itself an object? Objects have shape. What is
the shape of space? What is the shape of the physical universe. What is the shape of the
Platonic universe?
If a person (a thought, feeling and symbol object) were to read this section with the
hope of nding a clear denition of object, s/he might be disappointed. Yet, Heinz von
Foerster has suggested the enticing notion that objects are tokens for eigenbehaviors.
There is a behavior between the perceiver and the object perceived and a stability or
repetition that arises between them. It is this stability that constitutes the object
(and the perceiver). In this view, one does not really have any separate objects, objects
are always objects perceived, and the perceiver and the perceived arise together in
the condition of observation.
2.1 Compresence and coalescence
The world appears to be the union of separate objects, each a union of ever-smaller
particles until this reduction recedes beyond our ability to perceive differences. And yet,
at the same time, the world appears as a unity within which all these apparently
separated entities reside within the whole. For an observer these are two primary
modes of perception compresence and coalesence. Compresence connotes the
coexistence of separate entities together in one including space. Coalesence connotes
the one space holding, in perception, the observer and the observed, inseparable in an
unbroken wholeness. Coalesence is the constant condition of our awareness.
Coalesence is the world taken in simplicity. Compresence is the world taken through
the lter of language.
This distinction of compresence and coalesence drawn by Bortoft (1971) can act as a
compass in traversing the domains of object and reference. Eigenform is a rst step in
K 7137721/12/2004RAVICHANDRAN127841
K
34,1/2
132
toward a mathematical description of coalesence. In the world of eigenform, the
observer and the observed are one in a process that recursively gives rise to each.
3. Shaping a world
We identify the world in terms of how we shape it. We shape the world in response to
how it changes us. We change the world and the world changes us. Objects arise as
tokens of behavior that leads to seemingly unchanging forms. Forms are seen to be
unchanging through their invariance under our attempts to change, to shape them.
Can you conceive of an object independent of your ability to perceive it? I did not
say an object independent of your perception.
Let us assume that it is possible to talk of the tree in the forest where we are not. But
how are we to speak of that tree? One can say, the tree is there. What does this mean? It
means that there is a potentiality for that tree to appear in the event of the appearance
of a person such as myself or yourself in the place called that forest. What is the tree
doing when I am not in the forest?
I will never know, but I do know that it obediently becomes treeish and located
when I am there. The quotation marks are indications of objects dissolving into
relationships. Whenever I am present, the world (of everything that is the case) is
seen through the act of framing. I imagine a pure world, unframed. But this is the
world of all possibility. As soon as we enter the scene the world is ltered and
conformed to become the form that frame and brain have consolidated to say is
reality.
4. Heinzs eigenform model
Heinz created a model for thinking about object as token for eigenbehavior. This model
examines the result of a simple recursive process carried to its limit.
For example, see Figure 1. That is, each step in the process encloses the results of
the previous step within a box.
Then the innite concatenation of F upon itself is an innite nest of boxes as shown
in Figure 2.
Figure 1.
Figure 2.
K 7137721/12/2004RAVICHANDRAN127841
EigenForm
133
An innite nest of boxes is invariant under the addition of one more surrounding box.
Hence this innite nest of boxes is a xed point for the recursion. In other words, if X
denotes the innite nest of boxes, then
X = F(X):
The innite nest of boxes is one of the simplest eigenforms. In the process of
observation, we interact with ourselves and with the world to produce stabilities that
become the objects of our perception. These objects, like the innite nest of boxes, go
beyond the specic properties of the world in which we operate. They attain their
stability through this process of going beyond the immediate world. Furthermore, we
make an imaginative leap to complete such objects that are tokens for eigenbehaviors.
It is impossible to make an innite nest of boxes in the physical world. We do not make
it. We imagine it. In imagining that innite nest of boxes, we arrive at the eigenform,
which is the object for this process.
But does the innite nest of boxes exist? Certainly it does not exist in this page or
anywhere in the physical world with which the writer or presumably the reader is
familiar. The innite nest of boxes exists in the imagination!
Just so, an object in the world (cognitive, physical, ideal,. . .) provides a conceptual
center for the investigation of a skein of relationships. An object can have varying
degrees of existence just as does an eigenform. If we take Heinzs suggestion to heart,
that objects are tokens for eigenbehaviors, then an object in itself is a symbolic entity,
participating in a network of interactions, taking on its apparent solidity and stability
from these interactions. In and of itself each object is as imaginary as a pure eigenform.
Heinzs model can be expressed (as indeed he did express it) in quite abstract and
general terms. Suppose that we are given a recursion symbolically by the equation:
X(t -1) = F(X(t)):
here X(t) denotes the condition of observation at time t. X(t) could be as simple as a set
of nested boxes, or as complex as the entire conguration of your body in relation to
the known universe at time t. Then F(X(t)) denotes the result of applying the
operations symbolized by F to the condition at time t. You could, for simplicity,
assume that F is independent of time. Time independence of the recursion F will give
us simple answers and we can later discuss what will happen if the actions depend
upon the time. In the time independent case we can write
J = F(F(F(. . .)))
the innite concatenation of F upon itself. We then see that
F(J) = J
since adding one more F to the concatenation changes nothing. Thus J, the innite
concatenation of the operation upon itself, leads to a xed point for F. J is said to be the
eigenform for the recursion F. It is just like the nested boxes, and we see that every
recursion has an eigenform. Every recursion has a xed point.
Must the xed points of our recursions always live in a new domain? Certainly not.
For example, the number two is a xed point for f(x) = x
*
x 22; and the number
1 -Sqrt(2) is a xed point for F(x) = 2 -1=x: These numbers are part of the
K 7137721/12/2004RAVICHANDRAN127841
K
34,1/2
134
domain of real numbers usually assumed in working with numerical recursions. This
last example is worth comparing with the innite nest of boxes. If we ask for a xed
point for F(x) = 2 -1=x we are asking for an x such that x = 2 -1=x: Hence we
ask for x such that x
*
x = 2x -1; a solution to a quadratic equation. And one
veries that (1 -Sqrt(2))(1 -Sqrt(2)) = 2(1 -Sqrt(2)) -1: Hence x = 1 -
Sqrt(2) is an example of a xed point for F(x).
On the other hand, following the proof of the theorem, we nd that
J = F(F(F(. . .))) = 2 -1=(2 -1=(2 -1=(2 - )));
an innite continued fraction that formally satises the equation J = F(J): In this case,
we can make numerical sense of the innite construction. In general, we are challenged
to nd a context in which the innite concatenation of the operator makes sense.
The place where this sort of construction reaches a conceptual boundary is
met in dealing with all solutions to a quadratic equation. There we can begin
with the equation x
*
x = ax -b with roots x = (a -Sqrt(a
*
a -4b))=2 and
x = (a 2Sqrt(a
*
a -4b))=2: If (a
*
a -4b) , 0 then the roots are imaginary.
On the other hand, we can rewrite the quadratic (dividing by x for x not zero) as
x = a -b=x = f(x):
Associating to this form of the quadratic the eigenform
E = f(f(f(f(. . .))));
we have
E = a -1=(b -1=(a -1=(b - ))) with f(E) = E:
Thus, E is a formal solution to the quadratic equation, and the consecutive terms
E1 = a; E2 = a -1=b; E3 = a -1=(b -1=a); . . .
will converge to one of the roots when the roots are real, but will oscillate with no
convergence when the roots are imaginary. Nevertheless, this series and its associated
eigenform are very closely related to the complex solutions, and the eigenform provides
a conceptual center for the investigation of these relationships (Kauffman 1987, 1994).
We end this section with one more example. This is the eigenform of the Koch
fractal (Kauffman, 1987). In this case, one can symbolically write the eigenform
equation
K = K{K K}K
to indicate that the Koch Fractal re-enters its own indicational space four times (i.e. it is
made up of four copies of itself, each one-third the size of the original). The curly
brackets in the center of this equation refer to the fact that the two middle copies within
the fractal are inclined with respect to one another and with respect to the two outer
copies. Figure 3 shows the geometric conguration of the re-entry.
In the geometric recursion, each line segment at a given stage is replaced by four
line segments of one-third its length, arranged according to the pattern of re-entry as
shown in Figure 3. The recursion corresponding to the Koch eigenform is shown in
Figure 4. Here we see the sequence of approximations leading to the innite
self-reecting eigenform that is known as the Koch snowake fractal.
K 7137721/12/2004RAVICHANDRAN127841
EigenForm
135
Five stages of recursion are shown. To the eye, the last stage vividly illustrates how the
ideal fractal form contains four copies of itself, each one-third the size of the whole. The
abstract schema
K = K{K K}K
for this fractal itself can be iterated to produce a skeleton of the geometric recursion:
Figure 3.
Geometric conguration of
the re-entry
Figure 4.
Recursion corresponding
to the Koch eigenform
which leads to the innite
self-reecting eigenform
(Koch snowake fractal)
K 7137721/12/2004RAVICHANDRAN127841
K
34,1/2
136
K = K{K K}K = K{K K}K{K {K K} K K {K K} K} K{K K}K =
We have only performed one line of this skeletal recursion. There are 16 Ks in this
second expression just as there are 16 line segments in the second stage of the
geometric recursion. Comparison with this abstract symbolic recursion shows how
geometry aids the intuition. The interaction of eigenforms with the geometry of
physical, mental, symbolic and spiritual landscapes is an entire subject that is in need
of deep exploration. Compare with Kauffman (1987).
It is usually thought that the miracle of recognition of an object arises in some
simple way from the assumed existence of the object and the action of our perceiving
systems. What is to be appreciated is that this is a ne tuning to the point where the
action of the perceiver, and the perception of the object are indistinguishable. Such
tuning requires an intermixing of the perceiver and the perceived that goes beyond
description. Yet, in the mathematical levels, such as number or fractal pattern, part of
the process is slowed down to the point where we can begin to apprehend it. There is a
stability in the comparison, in the one-to-one correspondence that is a process
happening at once in the present time. The closed loop of perception occurs in the
eternity of the present individual time. Each such process depends upon linked and
ongoing eigenbehaviors and yet is seen as simple by the perceiving mind.
5. Eigenform and eigenvector quantum mechanics
In quantum mechanics, observation is modeled not by the eigenform, but by its
mathematical relative the eigenvector. An eigenvector v is a non-zero vector such that
Hv = kv for some operator observable H and constant k. The constant k is what is
regarded as the quantity that is observed (for example, the energy of an electron).
The operator His taken to be a linear operator on a vector space. The vector space is an
innite dimensional Hilbert space. These are particular properties of the mathematical
context of quantum mechanics. The k can be eliminated by replacing H by G = H=k
(when k is nonzero) so that Gv = (H=k)v = (Hv)k = kv=k = v. In quantum
mechanics observation is founded on the production of eigenvectors v with Gv = v
where v is a vector in a Hilbert space and G a Hermitian linear operator on that space.
Heinz was certainly aware, as a practicing physicist, of this model of observation in
quantum theory. His theory of eigenforms is a sweeping generalization of quantum
mechanics that creates a context for understanding the remarkable effectiveness of
that theory. If indeed the world of objects that we take to be an objective (sic) reality is,
in fact, a world of tokens for eigenbehaviors, and if physics demands forms of
observations that give numerical results, then the simplest example of such
observation is the Hermitian observable in the quantum mechanical model.
This is a reversal of epistemology, a complete turning of the world upside down.
Heinz has tricked us into considering the world of our experience and nding that it is
our world, generated by our actions and that it has become objective through the
self-generated stabilities of those actions. He has convinced us to come along with him,
and see that all of cybernetics conrms this point of view. He has left the corollaries
to us. He has not confronted the physicists and the philosophers head on. He has
brought us into his world and let us participate in the making of it. And he has pointed
to the genesis and tautological nature of quantum theory to those of us who might ask
the question.
K 7137721/12/2004RAVICHANDRAN127841
EigenForm
137
But he has also left the consequence of the question to us. For if the world is a world
of eigenforms and most of them are in time oscillatory, and unstable, must we insist on
stability at the level of our present perception of that world? In principle, there is an
eigenform, but that form leads always outward into larger worlds and new
understanding. In the case of quantum mechanics, the whole theory has the appearance
of an elementary exercise, conrming the view point of objects as tokens for
eigenbehaviors in a special case. Heinz leaves us with the conundrum of nding the
more general physical theory that conrms that special case.
This dilemma is itself a special case of the dilemma that Heinz has given us. He said
it himself many times. If you give a person an undecideable problem, the action of that
person in attempting to solve the problem shows who is that person and what is the
nature of his/her creativity.
6. A conversation with Ranulph Glanville
This essay has its beginnings in a conversation with Ranulph Glanville. Ranulph
asked Does every recursion have a xed point?, hoping for a mathematicians
answer. And I said rst, Well no, clearly not, after all it is common for processes to go
into oscillation and so never come to rest. And then I said, On the other hand, here is
the theorem:
Theorem. Every recursion has a xed point.
Proof. Let the recursion be given by an equation of the form
X
/
= F(X)
where X
/
denotes the next value of X and F encapsulates the function or rule that
brings the recursion to its next step. Here F and X can be any descriptors of actor and
actant that are relevant to the recursion being studied. Now form
J = F(F(F(F(. . .))));
the innite concatenation of F upon itself.
Then, we see that
F(J) = F(F(F(F(F(. . .))))) = J:
Hence, J is a xed point for the recursion and we have proved that every recursion has
a xed point. QED A
Ranulph said Oh yes I remember that! Can I quote your proof?, and I said
Certainly, but you will have to make your attribution to Heinz and his paper Objects:
Tokens for (Eigen-)Behaviors (von Foerster, 1981b, pp. 274-85), for that is where I
came to appreciate this result, although I rst understood it via the book Laws of
Form (Spencer-Brown, 1969).
And I went on to say that this theorem was in my view a startling magicians trick
on Heinzs part, throwing us into the certainty of an eigenform (xed point)
corresponding to any process and at the same time challenging us to understand the
nature of that xed point in some context that is actually relevant to the original
ground of conversation. Ranulph agreed, and our e-mails settled back into the usual
background hum.
K 7137721/12/2004RAVICHANDRAN127841
K
34,1/2
138
6.1 Audio-activity and the social context
I kept thinking about that question, and wondering about nding a good mathematical
example. Then I remembered learning about the audio-active sequence of numbers
from Conway (1985). This is a number sequence that begins as:
1; 11; 21; 1211; 111221; 312211; 13112221; 1113213211; . . .
Can you nd the next number in the sequence? If you read them out loud, the
generating idea becomes apparent
one; one one; two ones; one two; one one; . . .
Each term in the sequence is a description of the digits in the previous member of the
sequence. The recursion goes back and forth between number and description of
number. What happens as this recursion goes on and on?
Here is a bit more of it:
1
11
21
1211
111221
312211
13112221
1113213211
31131211131221
13211311123113112211
11131221133112132113212221
31132221232112111312211312113211
13211332111213122112311311222113111221131221
Now you can begin to see that there is a approach to a triple of innite sequences, each
describing the next, with the rst describing the last. This triple is the limiting
condition of the audio-active sequence. In one sense the audio-active sequence oscillates
among these three sequences (in the limit), and yet in another sense this triplet of
innite sequences is the eigenform in back of the audio-activity!
A = 11131221131211132221. . .
B = 3113112221131112311332. . .
C = 132113213221133112132123. . .
The triple of innite sequences are built by continually cycling the self-description
through the three sequences. This leads to a denite and highly unpredictable buildup
of the three innite sequences, A, B, and C such that Bdescribes A, C describes B and
A describes C! (Figure 5).
This triplication is the eigenform for the recursion of the audio-active sequence. The
triplicate mutual description is the xed point of this recursion. With this example,
we begin to see the subtlety of the concept of an eigenform, and how it may apply to
diverse human situations. For indeed imagine the plight of three individual human
K 7137721/12/2004RAVICHANDRAN127841
EigenForm
139
beings Alice, Bob and Carol who each take on the task to describe another, with Bob
describing Alice, Carol describing Bob and Alice describing Carol. In the mutual round
of their descriptions they may converge on a mutual agreement as do the triplet of
audio-active sequences (in the limit). Yet, it may take some coaxing to bring forth the
agreement and some creativity as well. More complex social situations will be beyond
calculation, and yet, the principles of the interaction, the possibility of eigenforms will
apply. The concept is powerful and important to consider, particularly when one is
faced with the incalculable nature of complex interaction.
7. Generation of objects
The true question about an object is: How is it generated?
The false question about an object is: What is its classication?
Take a mathematical case in point. Let R be the set of all sets that are not members
of themselves. (Russells famous paradoxical set.) We symbolize R as follows.
Let AB denote the condition that B is a member of A.
Dene R by the equation
RX = ,XX
which says X is a member of R means that it is not the case that X is a member of X.
From this we reach the paradox at once. Substitute R for X you obtain:
RR = ,RR
R is a member of R means that it is not the case that R is a member of R.
Something curious has happened. We attempt to classify R by nding if it was or
was not a member of itself and we are led into a round robin that oscillates between
membership and nonmembership. Classication creates trouble.
Ask how R is generated.
We start with some sets we know. For example, the empty set is not a member of
itself, neither is the set of all cats. So a rst approximation to R could be
R1 = { { }; Cats};
where Cats denote the set of all cats (Cats is not a cat.). Now we note that R1 is also
not a member of itself. So we have to add R1 to get a better approximation R2.
R2 = { { }; Cats; { { }; Cats} }:
But R2 is also not a member of itself and so we would have to add R2 and keep on with
this as well as throwing in other sets that come along and are normal. A set is normal if
it is not a member of itself.
Figure 5.
K 7137721/12/2004RAVICHANDRAN127841
K
34,1/2
140
This process will never end! Suppose S is any collection of sets that are not members of
themselves, then S itself is not a member of itself. S is barred from self-membership by
the condition that members of S must not be members of themselves.
But this means that S CANNOT be the set of ALL sets that are not members of
themselves, since S itself is a set that is not a member of itself, and S is not in S!
So we have found that any attempt to create the set of ALL normal sets is doomed to
failure. It is always possible to make new normal sets from the ones we already have.
This is good news. Normal sets are creative in this way. By taking a look at how the
Russell set could be generated, we nd that the Russell set will never be completed.
We nd that it is possible to have sets that continue to evolve in time and that time is a
necessary concept in mathematics. Only in the course of time can we conceive of
a Russell set. The set R and the eigenform RR exist only as instructions for action.
They do not exist as completed entities.
8. Number and multiplicity
McCulloch (1965) wrote What is a number that a man may know it, and what is a man
that he may know a number? It is a mutual shaping that produces these numbers,
these ideal objects made from thin air and notches on a stick, and sheep in the eld and
stars in the sky, and all the discrete appearances of object and relation that ll our time
and space.
I create distinct numbers by the conventions of sets and frames.
0 = { }
1 = { { } }
2 = { { }; { { } } }
3 = { { }; { { } }; { { }; { { } } } }
. . .
An innite sequence of distinct objects, each carrying its own multiplicity. Are each of
these also to be seen as eigenforms? Should not each and every one of my constructions be
seen as a formin the limit of recursions? Why do natural objects appear full born and just
there, when I amsure that they are all limits of recursive process? AmI daft to consider it?
Is not the verytexture of reective thought a limit of such a process? Must we go to innity
to nd the eigenforms. I do not want to fall into innite repetition. Not again.
9. Russells numbers
Turn now to Bertrand Russells denition of number: The number of a class is the class
of all classes in one-to-one correspondence with that class.
Recall that two classes (sets) S and T are said to be in one-to-one correspondence if
there is a mapping f: S--T that is one-to-one and onto, i.e. every t in T is of the form
f(s) for some s in S, and f(s) = f(s
/
) only when s and s
/
are equal in S. We could say
that S is a operator or rule(r) that can be used to measure the size of T, and that S and
Tare in one-to-one correspondence exactly when the ruler ts Tperfectly. In this sense
the number 1 = {{ }} is a measuring rod for singletons, the number 2 = {{ }; {{ }}} a
measuring rod for pairs.
K 7137721/12/2004RAVICHANDRAN127841
EigenForm
141
We can write n(X) = X if it is possible to establish a one-to-one correspondence
between n and X. As far as number is concerned 3({a; b; c}) = {a; b; c}: In this way
each number becomes an operator and the numerosity of a collection is the property of
that collection to be an eigenform for some number n.
In Russells denition of number we see the codication of that relationship between a
person and his world that brings forth objects through the actions of the person in
relationship to the stabilities of the world. These actions give rise to the higher eigenforms
that we call numbers. The stabilities themselves, we tend to relegate to the physical realm,
yet they include the stabilities of cognition the constancy of objects and relationships in
the visual eld, the balance of body, dexterity, gesture and imagined surrounding world.
We take these for granted, and it is usually thought that the miracle of recognition of
an object arises in some simple way from the assumed existence of the object and the
action of our perceiving systems. What is to be appreciated is that this is a ne tuning to
the point where the action of the perceiver and the perception of the object are
indistinguishable. Such tuning requires an intermixing of the perceiver and the
perceived that goes beyond description. In the mathematical levels, such as number, part
of the process is slowed down to the point where we can begin to apprehend the process.
There is a stability in the comparison, in the one-to-one correspondence that is a process
happening at once in the present time. It is this stability, this eigenform that constitutes
my knowledge of two hands as two groups of ngers. Each such process depends upon
linked and ongoing eigenbehaviors and yet is seen as simple by the perceiving mind.
10. Church and Curry
Church and Curry (Barendregt, 1981) showed (in the 1930s, long before Heinz wrote
his essays) how to make eigenforms without apparent excursion to innity.
Their formalism is usually called the lambda calculus.
Here is how it works.
We wish to nd the eigenform for F. We want to nd a J so that F(J) = J: Church
and Curry admonish us to create an operator G with the property that
G(X) = F(X(X))
for any X. That is, when G operates on X, G makes a duplicate of X and allows X to
act on its duplicate. Now comes the kicker.
Let G act on herself and look!
G(G) = F(G(G))
So G(G), without further ado, is a xed point for F. We have solved the problem
without the customary ritual excursion to innity.
I like to call the construction of the intermediate operator G, the gremlin
(Kauffman, 1995, 2001b). Gremlins seem innocent enough. They duplicate entities that
meet, and set up an operation of the duplicate on the duplicand. But when you let a
gremlin meet a gremlin then strange things happen. It is a bit like the story of the
sorcerers apprentice, except that here the sorcerer is the mathematician or computer
scientist who controls context, and the gremlins are like the self-duplicating brooms in
the story. The gremlins can go wild without some control. In computer science the
gremlins are programs with loops in them. If you do not put restrictions on the loops,
things can get very chaotic!
K 7137721/12/2004RAVICHANDRAN127841
K
34,1/2
142
Once an appropriate gremlin is in place, clocks will tick and numbers will count.
Here is a simple model.
Please agree that @ = @* so that @ is a xed point for the token
*
.
Then
@ = @* = @** = @*** = @**** = @***** =
A counting process ensues immediately from the equation
@ = @*:
That equation is the result of the gremlin XG = XX* and @ = GG:
Counting is the emanation of self-reference.
Counting is the unfolding of an eigenform.
Here we see how close are the concepts of process (counting) and xed point
(eigenform). It is through the eigenform that counting occurs, but we still need a higher
level of observation to make use of this capacity. It is still necessary to be able to read,
start and stop the clock. This points to the fact that a formal eigenform must be placed
in a context in order for it to have human meaning. The struggle on the mathematical
side (or computer science side) is to control recursions, bending them to desired ends
(Kauffman, 2001a) for a discussion of these concepts in the context of proofs generated
by humans and by machines. The struggle on the human side is to cognise a world
sensibly and communicate well and effectively with others. For each of us, there is a
continual manufacture of eigenforms (tokens for eigenbehavior) but such tokens will
not pass as the currency of communication unless we achieve mutuality as well. One
can say that mutuality itself is a higher eigenform. Achieving mutual understanding
will be recognized when we have it or begin to have it. As with all eigenforms, the
abstract version exists. Realization can happen in the course of time.
11. Differentiation creates number
What does this Church Curry method have to do with the invariance of the exponential
function under differentiation?
D(exp(t)) = exp(t) where D = d=dt?
In fact,
exp(t) = a[0] -a[1] -a[2] - _ c
where a[n] = (t ^ n)=n! so that
Da[0] = 0;
Da[n -1] = a[n]
from which it follows that
D(exp(t)) = exp(t):
Note that this is like modeling D[X] as the operator that REMOVES a box from around
X. Strange thoughts ensue such as those shown in Figure 6.
K 7137721/12/2004RAVICHANDRAN127841
EigenForm
143
The strangeness here is the meaning of the innite concatenation of the operator D. For
if D* = . . .DDD; then D
*
(J) = Nothing in at least one interpretation of these
symbols.
Let us return to the construction of numbers.
0 = { }
1 = { { } }
2 = { { }; { { } } }
It is a recursive process where
0 = { }
and
n -1 = {0; 1; 2; 3; . . . ; n} = {Dn; n}
where DS is the list obtained from a set S by removing its outer bracket.
Note that D{ } is nothing, D{ { } } = { }: D{ { }; { { } } } = { }; { { } } and so on.
For the limit singleton W = {W} we have DW = W: We now have the basic
equation
N-1 = {DN; N}:
In this form, the counting process resists the production of xed points. For example, if
we let
Figure 6.
Illustrations of strange
thoughts using the
operator D
K 7137721/12/2004RAVICHANDRAN127841
K
34,1/2
144
I = {0; 1; 2; 3; . . .}
be the rst ordered countable innity of integers, then
I -1 = {DI; I} = {0; 1; 2; 3; . . . ; I}
is a new set distinct from I, and I - 2 is distinct from I - 1. The counting process
continues without end.
12. The object of set theory
Let us lookat objects fromthe point of viewof a set theoretician. If AandBare objects, then
we can forma newobject C = {A; B}; the set consisting of Aand B. This seems harmless
enough. After all, if Chicago and NewYorkare objects, then the set of large coastal cities in
the United States should also be an object, albeit of a different type. We give up something
with these mathematical objects. We do not assume that they have specic spatial
locations. After all, what is the spatial location of the set {Chicago, New York}?
Take New York. This is a good big object to talk about. It is a place. It has a location.
It has contents, all the people in it, all the goods and people and ideas and music
running through it.
And we will leave all that and just take the set theoretic point of view and look at the
singleton set {New York}. Now {New York} is not New York. Not by a long shot! New
York is a hustling bustling metropolis on the East Coast of the United States of
America. New York has millions of inhabitants and buildings, and New York is
constantly changing. On the other hand, the singleton {New York} has exactly one
member. It never changes. It is always the set whose member is New York. On top of
this, once we have admitted the singleton object {New York} into existence, we are
compelled to allow the singleton of its singleton to come on the stage with its only
member the singleton of New York: {{New York}}. There is a innity of singleton
objects derived from New York waiting in the wings:
{New York}
{{New York}}
{{{New York}}}
{{{{New York}}}
. . .
{{{{{{{{{{{{{{{{{New York}}}}}}}}}}}}}}}}}
. . .
. . .
{{{{{{{{{{{{{{{{{{{{{{{ . . . }}}}}}}}}}}}}}}}}}}}}}} = W
The limiting New York singleton W has New York innitely down in the nest of
parentheses. New York has disappeared, and all that is left of the Cheshire Cat is its
grin. All that is left in the limit W is the fact that W is invariant under the act of
forming the singleton. We have that W is its own singleton!
W = {W}
Adding one more level of parentheses makes no difference. W is at a level where the
level and the metalevel are one. W is both object and subject of the set theoretic
discourse. And if you think that W has nothing to do with New York you are wrong!
K 7137721/12/2004RAVICHANDRAN127841
EigenForm
145
W is the very identity of New York. W is the ultimate singleton associated with
New York. Wis the essence of New York. And at the same time, Wis entirely content
free and has nothing to do with New York. W is just an innite nest of brackets.
An uninterpreted bit of self-reference in the void. Will you have it both ways? You
could locate W anywhere. Why not New York?
13. Singletons and eigenforms
The example of New York illustrates the extreme eigenform associated with any object
in the set theoretic universe. We can iterate the operation of framing X to form the
singleton {X} ad innitum, and lose X in the innite depths of the recursion. We lose
Xand regain the ubiquitous and self-referential W = {W}: This could bring one to be
suspicious of the concept of singleton set. After all, why should New York or any
proper object in the world be surrounded by an innite halo of singletons? When I eat
an apple, must I devour {apple}, {{apple}}, {{{apple}}},. . . as well? Quine and others
have suggested that we take a different approach to framing so that singletons do not
appear. One way to achieve this is to legislate that {S} = S for any S that is
non-empty.
Think about this proposal. We would have {{ }} = { } and there would be no way
to produce a set with one element! One could say that there are special objects in the
theory, let us call them a,b,c,. . . such that each object is its own singleton:
{a} = a; {b} = b; {c} = c; . . .
Then, at least the singletons for these real objects collapse back to them. This
approach raises many questions. What are the special objects. Certainly no
mathematician would want the empty set to be among them, since we wish to
discriminate between the empty set and the set whose member is the empty set.
Searching for these special objects is something like searching for elementary particles.
Where are they? Could it be that I am a special object? Let us see. Is it the case that
I = {I}: Why yes indeed! I can frame (think about) myself and I am still myself! In fact
if we interpret the emergence of a frame as an act of reection (thought), then the
special objects appear as elements of I-ness, as signs for themselves in the sense of
Charles Sanders Peirce (Kauffman, 2001b).
But set theory goes its own way, and would weave these special objects together
into hierarchies that embody singletons once again. Take two specials a and b.
Form their union {a, b} Is this special? Why not? Why not allow that if any element of
a set S is special, then S is also special? This still allows room for classical
mathematics. We can always form sets like {{ },{{ }}} that are not special. Note that
a = {a} does not imply that a is equal to an innite nest of parentheses. The innite
nest W = {{{{{{{ . . . }}}}}}} is but one of many special objects that are their own
singletons.
By including special objects into set theory and these rules for their composition, we
have created a model for a set theoretic world that contains a parable (a parallel parable!)
of the mental and the physical. The purely mental world is the class of sets generated
from the empty set. The purely physical world is the class of sets generated from the
special objects. The interface of mental and physical occurs as they touch in the limit of
nested parentheses. W = {{{{{{{ . . . }}}}}}} is an amphibian living in both worlds.
W is the eigenform that crosses the boundary from the mental to the physical.
K 7137721/12/2004RAVICHANDRAN127841
K
34,1/2
146
14. The form of names
The simplicity of a thought, the apparent clarity of distinction is mirrored in the sort of
eigenforms that come from the Church-Curry realm as described in the last section.
Consider a linguistic example. Each person has a name (at least one). In the course of
time we are introduced to people and come to know their names. We know that name
not as an item to look up about the person (and this applies to certain objects as well)
but as a direct property of the person. That is, if I meet Heinz he appears to me as
Heinz, not as this person with certain characteristics, whose name I can nd in my
social database if I care to do so. It is like this only when we are rst introduced.
At the point of introduction there is this person and there is his name separate from
him. Once learned, the name is shifted and occurs in space right along with the person.
Heinz and his name are in the same cognitive space which is also in the same place as
the apparent physical space. We can observe this shifting process in the course of
learning a name. We can also observe how physical and cognitive spaces are
superimposed. The many classical optical illusions illustrate these matters vividly.
Now we have Heinz with his name inseparable from his presence, and this is true
even if he is not physically present, for the shift has occurred and will not be undone.
But we also have his name Heinz separate from him, and able to be pinned upon
another. And we have his name not quite separate from him, but rather this Heinz is
the name of the name we have attached to him! This is Heinzs metaname. How do we
distinguish among all these different names for Heinz? We use the same symbols for
them, yet they are different. Let us choose a way to indicate the differences. We start
with the reference.
Heinz --Cybernetic Magician
(The arrow will indicate that the entity on the left is the name of the entity on the right.)
We get to know him and shift the reference.
#Heinz --Cybernetic Magician Heinz
Now the name is in the cognitive space of Heinz, and the metaname #Heinz refers to
that conjunction. We shall call this the indicative shift.
name --object
#name --object name
The indicative shift occurs, constantly weaving the apparent external reality with the
linguistic reality.
Self-reference occurs when one calls up (names) the metanaming operator.
At rst the metanaming operator is not marked and no name has been chosen for it.
But then its name is chosen (as #). We have:
# --
That is, # refers at rst to the singular place where there is an absence of naming, a
void in the realm of distinctions.
Then the shift occurs. We have the reference of the meta-naming operator to itself
(as the operator enters a space formerly void!).
K 7137721/12/2004RAVICHANDRAN127841
EigenForm
147
#--#
Suppose that the meta naming operator has another name, say M. Then we have
M--#
which shifts to a self-reference at the second articulated level of meta naming.
#M--#M
These are the eigenforms of self consciousness in the realm of names.
Von Foerster (1981c) said:
I am the observed link between myself and observing myself.
Self-reference, the action of a domain upon itself, leading to cognition, is the beginning
of the realm of eigenforms in Heinzs world. I am I is the shortest explicit loop. I am
is the shortest prescription for eigenform.
Remark. The indicative shift provides the formalism for statements that refer to
their own referents. Let
g --F#
denote a reference of g to the statement (form) F#. Shifting, we obtain
#g --F#g:
This is the form of F speaking about #g, where #g is the name of the statement F#g.
Thus, F#g is speaking about itself. For a longer discussion of this structure,
see (Kauffman, 1994, 1995).
15. Cantors diagonal argument, Lawveres eigenform and imaginary
values beyond true and false
We cannot resist including here a magnicent application of xed points as eigenforms
that is due to the category theorist (Lawvere, 1970). Lawvere found a new way to prove
a famous theorem of Georg Cantor, founder of transnite set theory. Cantors theorem
states the following.
Cantors theorem. For every set S, there are more subsets of S than there are
members of S.
Lawveres Proof of Cantors Theorem. Every subset A of a set S can be specied
by a mapping from S to the set of truth values {T,F}, specifying T if an element is in
the subset A and specifying F if the element is not in the subset. We can write
kAl : S--{T; F}
for this function.
Now suppose that there are just as many subsets of S as there are elements of S.
Then we could label each subset of S with an element of S. We would have subsets
indexed by elements of S, writing A[s] for the subset that corresponds to the element s.
We are going to prove that we always run out of labels! The crux of the matter is the
fact that the element s either is or is not a member of A[s].
The hypothesis that every subset has a label means that every corresponding
function kA[s]l(t) = True or False is really a function of two elements of S. We can
dene a function
K 7137721/12/2004RAVICHANDRAN127841
K
34,1/2
148
G : S S--{T; F}
(S S denotes pairs (s, t) of elements of S). By the formula
G(s; t) = kA[s]l(t);
and this function will specify T or F according to whether t is a member of the subset
with label s. Our assumption (which we want to show is absurd) is that there exists this
function G from S S to the set of truth values {T, F}.
Now Lawvere does the following. He shows that the labeling hypothesis leads to the
existence of a xed point for every mapping from the truth set {T, F} to itself! This is a
contradiction since negation (,T = F and ,F = T) has no xed points. Thus, the
existence of xed points is at the heart of Lawveres proof of Cantors Theorem.
How does Lawvere accomplish this feat?
Take any function J dened on the set of truth values to itself.
J : {T; F} --{T; F}:
Now look at the function
Q(t) = J(G(t; t)):
This is a function fromS to {T, F}, and so by our assumption there must be an element
s of S such that Q(t) = G(s; t) for all t in S. Thus, we have
J(G(t; t)) = G(s; t) for all t:
Hence,
J(G(s; s)) = G(s; s):
where G(s, s) is xed by the mapping J.
We have shown that every mapping from {T, F} to itself has a xed point! This is
the desired contradiction, and completes Lawveres proof of Cantors Theorem. A
If we had enlarged the truth set to
{T; F; I}
where ,I = I is an eigenform for negation, then G(s, s) would have value I. What does
this mean? It means that the index of the set A corresponding to G would have an
oscillating truth value. If that index s is in A then it is not in A and if the index is not
in A then it is in A. We would be propelled into sets that vary in time. Not a bad idea,
even if it is time to end this essay.
16. Epilogue
This paper has been a contemplation of the concept of recursion in the context of
second-order cybernetics. The simple idea of iterating an operation upon itself is seen
to be a key to understanding the nature of objects and the relationship of an observer
and the apparent world of the observer. In this view, the observer does not stand
outside the world and see it. Rather, what is seen is a token, an eigenform, of the
recursive participation of the observer in a world where there is no separation of
the observer and the observed. The experience of separation can just as well be
K 7137721/12/2004RAVICHANDRAN127841
EigenForm
149
an experience of joining in that participation. Objects become our own creations and
the world is the theatre of our actions upon it, which is us. It was Heinz von Foersters
ability to create himself, us, and our world as magic theatre that makes these views so
vivid to us. The existence of clear formalisms for recreating these ideas will help them
grow forward into the future. We imagine eternity, and in so doing create the ow of
time.
References
Barendregt, H.P. (1981, 1985), The Lambda Calculus Its Syntax and Semantics, North Holland,
Amsterdam.
Bortoft, H. (1971), The whole counterfeit and authentic, Systematics, Vol. 9 No. 2, pp. 43-73.
Conway, J.H. (1985), Private conversation.
Kauffman, L.H. (1987), Self-reference and recursive forms, Journal of Social and Biological
Structures, pp. 53-72.
Kauffman, L.H. (Ed.) (1994), Knot logic, Knots and Applications, World Scientic, Singapore,
pp. 1-110.
Kauffman, L.H. (1995), Virtual logic, Systems Research, Vol. 13, (Festschrift for Heinz von
Foerster), No. 3, pp. 283-310 (1996) (A short version appears in 10th IEEE Intl. Symposium
on Intelligent Control, Adrem Inc. Pub., pp. 89-98).
Kauffman, L.H. (2001a), The Robins problem computer proofs and human proofs, Kybernetes,
Vol. 30 No. 5/6, pp. 726-51.
Kauffman, L.H. (2001b), The mathematics of Charles Sanders Peirce, Cybernetics and Human
Knowing, Vol. 8 No. 1-2, pp. 79-110.
Kauffman, L.H. (2003), Eigenforms objects as tokens for eigenbehaviours, Cybernetics and
Human Knowing, Vol. 10 No. 3-4, pp. 73-89.
Lawvere, F.W. (1970), Introduction, Toposes, algebraic geometry and logic, Springer Lecture
Notes on Mathematics, Vol. 274, pp. 1-12.
McCulloch, W.S. (1965), What is a number that a man may know it, and a man, that he may
know a number?, Embodiments of Mind, MIT Press, Cambridge, MA, pp. 1-18.
Spencer-Brown, G. (1969), Laws of Form, George Allen and Unwin Ltd, Sydney.
von Foerster, H. (1981a), Observing Systems, The Systems Inquiry Series, Intersystems
Publications, Salinas, CA.
von Foerster, H. (1981b), Objects: tokens for (eigen-) behaviors, Observing Systems, The Systems
Inquiry Series, Intersystems Publications, Salinas, CA, pp. 274-85.
von Foerster, H. (1981c), Notes on an epistemology for living things, Observing Systems,
The Systems Inquiry Series, Intersystems Publications, Salinas, CA, pp. 258-71.
von Foerster, H. (1981d), On constructing a reality, Observing Systems, The Systems Inquiry
Series, Intersystems Publications, Salinas, CA, pp. 288-309.
K 7137721/12/2004RAVICHANDRAN127841
K
34,1/2
150